selan.aiGuide · Claude Code cost
Is a Claude Max plan cheaper than the API?
Only for a developer whose traffic would cost more than the plan at API list price. Max 5× is $100 a month and Max 20× is $200. At Claude Opus 5.5's list price, $200 buys 10 million output tokens or 50 million input tokens. Below that, the API is cheaper.
What does each option cost?
A Claude subscription is a flat fee per seat, paid whether the seat is used or not.
| Plan | Per seat, per month | What else to know |
|---|---|---|
| Pro | $20 | $200 a year if paid annually. |
| Max 5× | $100 | — |
| Max 20× | $200 | — |
| Team Standard | $25 | $240 a year if paid annually. |
| Team Premium | $125 | $1,200 a year if paid annually. Carries the same allowance as Max 5×. |
| Enterprise | $20 | Usage is billed per token on top of the seat. |
The API has no monthly fee. It bills every token, at a rate that depends on the model and on the kind of token.
| Model | Input | Output | Cache read | Cache write |
|---|---|---|---|---|
| Claude Fable 5.1 | $10 | $50 | $0.25 | $12.50 |
| Claude Opus 5.5 | $4 | $20 | $0.20 | $5 |
| Claude Sonnet 5.5 | $2 | $10 | $0.10 | $2.50 |
| Claude Haiku 5.5 | $0.10 | $0.50 | $0.01 | $0.125 |
Rates are in US dollars per million tokens. The cache write is the five-minute one.
Where is the break-even between Max and the API?
At however much traffic the monthly fee would buy at list price:
| $100 or $200 buys | On Opus 5.5 | On Sonnet 5.5 |
|---|---|---|
| Output tokens | 5M or 10M | 10M or 20M |
| Input tokens | 25M or 50M | 50M or 100M |
| Cache writes | 20M or 40M | 40M or 80M |
| Cache reads | 500M or 1,000M | 1,000M or 2,000M |
Real traffic mixes all four kinds, so no single row gives the answer. Multiply each kind of token by its own rate, add the four, and compare the month's total with the fee.
Why is the cheaper option not always the right one?
A seat cannot cost more than its fee, but it can stop.
Every Claude Pro, Max and Team seat is metered in a 5-hour window and a 7-day window, and a seat that fills either one waits for it to reset.
An API key has no window. It bills per token, and the only bound on its cost is the spending limit somebody sets on it. A job that must not stop halfway, such as a CI run, is safer on a key.
When each window resets, and what a seat shows once it fills one, is in what to do when Claude Code says rate limit reached.
How do you decide for a whole team?
Per provider, from each person's traffic rather than one figure for the company. A single total across providers hides the provider whose seats do not pay off.
Judge a seat over thirty days. It is charged by the month, and a verdict over seven days can flip by the thirtieth. If a seat costs more than its traffic would have cost at list price, cancel it.
How does Selan measure it on your own traffic?
Selan connects the seats and API keys your company already pays for, and records every request with the person, the account, the model and the cost. A seat's traffic is priced at what the same tokens would have cost per token, so seats and keys fit in one table.
For each provider, that table has the seats held, what they cost, the same traffic at list
price, your effective price per million tokens and the multiple between the two. The
verdict covers a fixed thirty-day window and either states the saving and the multiple or
says cancel. Every estimated figure carries a ~.
Requests use the seats first, and fall through to a key billed per token only when the seats are capped. A limit per person caps what each developer can spend on those keys.
The cost side is on the spend page, and how seats and keys are connected is in Providers. To set it up, start at the docs.
Free for up to three seats, and the Free plan needs no card.
The short version
- A seat is cheaper than the API only once a developer's traffic would cost more than its fee at list price.
- A seat can stop when it fills its 5-hour or 7-day window. An API key has no window, so a job that must not stop halfway is safer on a key.
- Decide per provider, from each person's traffic, and judge a seat over thirty days.
Plan prices are Anthropic's published list prices per seat. Token rates are the per-token list prices OpenRouter published on 13 October 2026, as Selan's price file read them. Both change: check the date before quoting a figure.